Correlation between multifidus muscle atrophy, spinopelvic parameters, and severity of deformity in patients with adult degenerative scoliosis: the parallelogram effect of LMA on the diagonal through the apical vertebra

Background There were several reports describing the biomechanics and microstructure of multifidus muscles in patients with lumbar disc herniation. However, correlations between lumbar multifidus muscle atrophy (LMA), spinopelvic parameters, and severity of adult degenerative scoliosis (ADS) have not been investigated. The study evaluated the impact of LMA and spinopelvic parameters on the severity of ADS. Methods One hundred and thirty-two patients with ADS were retrospectively reviewed. Standing whole-spine X-ray was used to evaluate the coronal (coronal Cobb angle, CA; coronal vertical axis, CVA) and sagittal (sagittal vertical axis, SVA; thoracic kyphosis, TK; lumbar lordosis, LL; pelvic incidence, PI; pelvic tilt, PT; sacral slope, SS) parameters. LMA was evaluated on axial T2-weighted magnetic resonance imaging (MRI) at intervertebral levels above and below the vertebra at the apex of the scoliotic curve. Clinical symptoms were evaluated by the Oswestry Disability Index (ODI) and the Japanese Orthopaedic Association (JOA) score. Multiple linear regression was used to assess correlations between LMA, spinopelvic parameters, and severity of scoliosis. Results LL and PT were negatively correlated with CA (P < 0.001); LL was positively correlated with SVA (P < 0.001). PI was positively correlated with CA (P < 0.001) and CVA (P < 0.001). PT (P < 0.001) and SS (P < 0.001) were negatively correlated with CVA. SS was negatively correlated with SVA (P < 0.001). Concave LMA at the upper or lower intervertebral level of the apical vertebra was positively correlated with CA (P ≤ 0.001); convex LMA at the upper or lower intervertebral level was negatively correlated with CA (P < 0.001). Convex LMA at the upper intervertebral level and concave LMA at the lower intervertebral level of the apical vertebra were negatively correlated with the SVA (P ≤ 0.001). At the upper intervertebral level, LMA on the concave side was positively correlated with CVA (P = 0.028); LMA on the convex side was negatively correlated with CVA (P = 0.012). PI was positively correlated with ODI (P < 0.001); PT (P < 0.001) and SS (P < 0.001) were negatively correlated with ODI. At the lower intervertebral level, LMA on the concave side was positively correlated with ODI (P = 0.038); LMA on the convex side was negatively correlated with ODI (P = 0.011). PI was positively correlated with JOA (P < 0.001); PT (P < 0.001) and SS (P < 0.001) were negatively correlated with JOA. Conclusions Spinopelvic parameters are correlated with the severity of ADS. Asymmetric LMA at both upper and lower intervertebral levels of the apical vertebra is positively correlated with CA. LMA on the diagonal through the apical vertebra is very important to maintain sagittal imbalance via parallelogram effect. LMA at lower intervertebral levels of the apical vertebra may have a predictive effect on ODI. JOA score seems to be more correlated with spinopelvic parameters than LMA.


(Continued from previous page)
Results: LL and PT were negatively correlated with CA (P < 0.001); LL was positively correlated with SVA (P < 0.001). PI was positively correlated with CA (P < 0.001) and CVA (P < 0.001). PT (P < 0.001) and SS (P < 0.001) were negatively correlated with CVA. SS was negatively correlated with SVA (P < 0.001). Concave LMA at the upper or lower intervertebral level of the apical vertebra was positively correlated with CA (P ≤ 0.001); convex LMA at the upper or lower intervertebral level was negatively correlated with CA (P < 0.001). Convex LMA at the upper intervertebral level and concave LMA at the lower intervertebral level of the apical vertebra were negatively correlated with the SVA (P ≤ 0.001). At the upper intervertebral level, LMA on the concave side was positively correlated with CVA (P = 0.028); LMA on the convex side was negatively correlated with CVA (P = 0.012). PI was positively correlated with ODI (P < 0.001); PT (P < 0.001) and SS (P < 0.001) were negatively correlated with ODI. At the lower intervertebral level, LMA on the concave side was positively correlated with ODI (P = 0.038); LMA on the convex side was negatively correlated with ODI (P = 0.011). PI was positively correlated with JOA (P < 0.001); PT (P < 0.001) and SS (P < 0.001) were negatively correlated with JOA. Conclusions: Spinopelvic parameters are correlated with the severity of ADS. Asymmetric LMA at both upper and lower intervertebral levels of the apical vertebra is positively correlated with CA. LMA on the diagonal through the apical vertebra is very important to maintain sagittal imbalance via parallelogram effect. LMA at lower intervertebral levels of the apical vertebra may have a predictive effect on ODI. JOA score seems to be more correlated with spinopelvic parameters than LMA.
Keywords: Adult degenerative scoliosis, Lumbar multifidus muscle atrophy, Spinopelvic parameters, Sagittal imbalance, Correlation Background Adult degenerative lumbar scoliosis (ADS) is defined as spinal deformity with a coronal deviation of greater than 10°in a skeletally mature patient, especially older than 40 years, without a history of scoliosis in childhood or adolescence [1]. The prevalence of ADS rises with age, with estimates ranging from 6 to 68%; therefore, it is becoming a major public health concern as the global incidence is increasing with the aging population [2,3].
There were several reports describing the biomechanics and microstructure of multifidus muscles in patients with lumbar disc herniation [4]. Results showed correlations between lumbar multifidus muscle atrophy (LMA) and chronic low back pain, disc degeneration, and radiculopathy [5,6]. However, correlations between LMA, spinopelvic parameters, and severity of ADS have not been investigated. Yagi et al. [7] reported that the cross-sectional area (CSA) of the multifidus (MF) and psoas (PS) were significantly smaller in degenerative lumbar scoliosis (DLS) patients. However, their muscle CSA analysis only included the L5-S1 level, which could not reflect the characteristics of paravertebral muscles around the apex vertebra. Our study simplified the approach to assessing LMA by using Goutallier Classification system [8]; LMA on the convex and concave sides of the scoliotic curve was measured on axial T2-weighted MR images at intervertebral levels above and below the apical vertebras. All of them would make the results more applicable.
The purpose of this study was to evaluate the correlation between LMA, spinopelvic parameters, and the severity of ADS and identify variables that predict progression in ADS. To our knowledge, this has not been previously reported.

Selection criteria
Patients with ADS that attended our inpatient clinic during the period from January 2016 to December 2017 were eligible for study. Inclusion criteria were age > 40 years at the time of attendance, medical records containing anteroposterior and lateral X-ray radiographs of total spine and magnetic resonance imaging (MRI) of the lumbar spine, and Cobb angle of lumbar curve in the coronal plane > 10°on a standing posteroanterior film. Exclusion criteria were history of scoliosis in childhood or adolescence, history of spinal surgery, local infection, inflammation around the spine, history of severe spinal trauma, spinal tumor, and presence of other systemic diseases that can affect spinal alignment (e.g., muscular dystrophy, ankylosing spondylitis, Parkinson disease). All patients included in our study provided written informed consent. This study has been approved by the institutional review board following the Declaration of Helsinki principles.

Evaluation of muscles in MRIs
In this study, 1.5 imaging system (Magnetom Symphony; Siemens, Berlin, Germany) for MRIs was used, and three T2-weighted axial images at intervertebral levels above and below the apical vertebras were obtained. The slices were separated by a 0.1-mm gap and were 4 mm thick; multifidus muscles on the convex and concave sides of the scoliotic curve were analyzed from the center slice of each of the three T2-weighted axial images. Muscle atrophy was known to be related to increased fatty infiltration; therefore, the Goutallier Classification system [8] was used to quantify muscle fatty degeneration in the lumbar multifidus muscle (Fig. 1).

Radiographic measurement and analysis
Standing whole-spine X-ray (Philips Digital Diagnost; Zhejiang Province, China) was used to evaluate the patients (Fig. 2). All radiologic parameters were measured twice at 1-month intervals by two researchers who were not involved in the patient encounters.

Clinical assessment
Demographic data, including patients' age and sex, were recorded. Clinical symptoms were evaluated by the Oswestry Disability Index (ODI) and the Japanese Orthopaedic Association (JOA) score. All of them were valid and rigorous functional measures used for assessing ADS.

Statistical analysis
Statistical analyses were performed using Statistical Package for the Social Sciences version 17.0 software (SPSS, Inc., Chicago, IL). Non-contiguous data was presented as number or ratio, whereas continuous variables were reported as mean ± standard deviations (SD). Shapiro-Wilk test was used to test for normality. Wilcoxon rank-sum test was used to compare the differences in LMA between the concave and convex sides at intervertebral levels above and below the apical vertebras, or between upper intervertebral levels and lower intervertebral levels of the apical vertebras in concave or convex sides. Pearson's correlation test was used to analyze the correlation between different parameters. Multiple linear regression was used to assess correlations between LMA, spinopelvic parameters, and severity of scoliosis. Statistical significance was set at P < 0.05.

Comparison of LMA between different intervertebral levels or sides
The Goutallier Classification system showed the following: at the upper intervertebral level, LMA (upper LMA, U-LMA) significantly increased on the concave side compared with the convex side (convex vs. concave, Z = − 7.616, P < 0.001); at the lower intervertebral level, LMA (lower LMA, L-LMA) similarly increased on the concave side compared with the convex side (convex vs. concave, Z = − 2.345, P = 0.019); on the concave side, no significant difference was found in between U-LMA and L-LMA (lower vs. upper, Z = − 0.093, P = 0.926); and on the convex side, U-LMA was significantly increased compared with L-LMA (lower vs. upper, Z = − 7.049, P < 0.001; Table 2; Fig. 3).

Correlation of influencing factors and severity of ADS
Multiple linear regression (R 2 = 0.705) of spinal pelvic parameters and CA showed there was a positive correlation between PI and CA (B = 1.519, P < 0.001); the correlations between PT (B = − 0.335, P < 0.001), LL (B = − 0.558, P < 0.001), and CA were negative (Table 4). On the concave side, both U-LMA (B = 4.266, P = 0.001) and L-LMA (B = 13.343, P < 0.001) were positively correlated with CA; on the convex side, both U-LMA (B = − 8.123, P < 0.001) and L-LMA (B = − 4.272, P < 0.001) were negatively correlated with CA; R 2 value of this model was 0.500 (Table 4). Table 5, which summarizes the correlation between spinal pelvic parameters, L-LMA, or U-LMA on concave or convex side and SVA, shows that LL was positively correlated with SVA (B = 0.342, P < 0.001) and SS was negatively  (Table 6). Multiple linear regression (R 2 = 0.680) of spinal pelvic parameters and ODI showed PI was positively correlated with is measured from the superior end plate of the cephalad end vertebra and the inferior end plate of the caudal end vertebra on the coronal plane; the coronal vertical axis (CVA) is defined as the distance from a perpendicular line drawn from the superior end plate of S1 vertebral body to the C7 plumb line on the coronal plane. b The sagittal vertical axis (SVA) is defined as the distance from a perpendicular line drawn from the superior end plate of S1 vertebral body to the C7 plumb line on lateral radiographs; the thoracic kyphosis (TK) is measured from the upper end plate of T2 to the lower end plate of T12; the lumbar lordosis (LL) is measured from the upper end plate of T12 to the upper end plate of S1; the sacral slope (SS) is defined as the angle between the cranial sacral end plate and a horizontal line; the pelvic incidence (PI) is defined as the angle subtend by a line connecting the center of the femoral head to the center of the cephalad end plate of S1 and a perpendicular line from the upper end plate of S1; the pelvic tilt (PT) is measured as the angle between the vertical axis and the line through the midpoint of the sacral end plate to femoral heads axis ODI (B = 0.407, P < 0.001); PT (B = − 0.229, P < 0.001) and SS (B = − 0.466, P < 0.001) were negatively correlated with ODI (Table 7). At the lower intervertebral level, LMA on the concave side was positively correlated with ODI (B = 0.735, P = 0.038); LMA on the convex side was negatively correlated with ODI (B = − 0.668, P = 0.011); however, this multiple linear regression model made little sense (R 2 = 0.051; Table 7). PI was positively correlated with JOA (B = 0.192, P < 0.001); PT (B = − 0.119, P < 0.001) and SS (B = − 0.213, P < 0.001) were negatively correlated with JOA; R 2 value of this model was 0.687 (Table 8). Duration of disease was positively correlated with SVA (B = 0.138, P < 0.001, R 2 = 0.111) and negatively correlated with CA (B = − 0.086, P = 0.017, R 2 = 0.043); no significant correlation was found between duration of disease and CVA (P = 0.838), ODI (P = 0.352), and JOA (P = 0.121). Considering the R 2 value of this model was very low, duration of disease was not a good predictor of the severity of ADS.

Discussion
The multifidus muscle is the most medially located back muscle; it is also the largest muscle, which spans the lumbosacral junction; in addition, it contributes to maintaining the erector posture of the trunk and to rotating and abducting the trunk [9]. Degeneration of soft tissue structures occurs in adult spinal degenerative disease [10]. Consequently, LMA may result in instability of the spine and exacerbate disc and facet degeneration in the lumbar spine [11]. Paraspinal muscle plays a more important role in maintaining the stability of L3-L4 segment than others [12]. This may explain why the apical vertebras of ADS patients most commonly occurred in L3 or L4 segment.
Hypotheses describing the mechanisms of LMA include disuse, denervation, inflammation, and injury [13][14][15][16]. In ADS patients, disuse and immobilization of the back muscles are common; these changes may cause atrophy at different intervertebral levels; furthermore, paraspinal denervation and re-innervation are common in disc herniation or nerve root compression [17]. Considering multifidus muscle is innervated by the dorsal root of the lumbar spinal nerve, atrophy of multifidus muscle innervated by medial branch of the dorsal ramus of the lumbar    nerve root would occur when the nerve root is compressed by herniated mass [18]. Sun et al. [17] stated that as there was no denervation phenomenon at the L3-L4 level, LMA could be the cause of disc degeneration; at the L5-S1 level, however, LMA could also be the consequence of L4-L5 disc herniation; pathogenetic mechanisms of U-LMA and L-LMA were different. Similarly, our study showed that LMA on the concave side was more severe than that on the convex side; on the convex side, there were differences in the causes of U-LMA and L-LMA. These indicated that the method, in which the axial image obtained at the level of the apex of the curvature was used as a reference for comparison to minimize the effect of the deformity on the morphometry of the paraspinal muscle, in the previous study might not be proper [19]. U-LMA may be the cause of ADS, while L-LMA may be the consequence of ADS. The positional change of the morphometry of paraspinal muscles will be influenced by the difference in the length of the arc in both sides of scoliosis; this is positively related to the radius of the arc; in addition, it is also proportional to their distance from the center of the axis of the spine. The CSA of paraspinal muscles can sometimes show discordant patterns of differences; the possible explanation is that the paraspinal muscles far from the axis of the spinal column will show the obvious effect of the positional change more than others that are closer to the axis of the spine; however, the influence of positional change on multifidus muscles may be caused by the change in the muscle itself, including the atrophy on the concave side or hypertrophy on the convex side [5,19]. Therefore, patients in supine position during the MRI examination will not influence the evaluation of LMA. Results of Pearson correlation analysis of spinopelvic parameters and LMA in our study showed that there were correlations among LL, SS, PI, and PT without TK. This is because spinopelvic parameters are geometrically related, such that PI is equal to the sum of SS and PT; variations in the lower arc of lordosis are determined by the sacral slope; when the sacral slope increases, the lower arc of lordosis increases, then the global curvature of lordosis increases as well; thoracic segments are supported by ribs with relatively poor compensatory ability; therefore, the correlation between TK and other radiographic parameters is not significant [20]. Acting like a bowstring, the multifidus muscle could switch compression loading to stretch loading and transmit some of the axial compression force on the disc to the anterior  longitudinal ligament, then maintains the spinopelvic parameters [21]. However, our study showed that L-LMA on both concave and convex sides were positively correlated with PT; U-LMA on the concave side was positively correlated with LL; L-LMA on the concave side was positively correlated with PI. These revealed that when the ADS occurred, bowstring effect of multifidus muscle would be influenced by multiple complex factors; the status of multifidus muscle on the concave side would play more important roles in maintaining spinopelvic parameters than others. A high pelvic incidence is associated with long, curved lumbar lordosis; this reciprocal association is an important component of overall sagittal alignment [22]. It was reported that lumbar hypolordosis was associated with lateral listhesis, vertebral rotation in ADS patients, which would aggravate scoliosis [23]. Therefore, in our study, LL and PT are both protective factors of CA; LL was positively correlated with SVA. PI reflects compensatory ability of the lumbar spine and pelvis in maintaining global alignment of the spine. In patients with high PI, the occurrence of ADS may mean a more severe decompensated state than others. In consequence, PI was positively correlated with CA and CVA in our study [24]. A previous study reported that there was a correlation between an anterior shift in the C7 plumb line and a vertically oriented sacrum [25]. In addition, the sacral slope was positional parameters that can be affected by changes in the alignment of the lower extremities, which would also influence CVA [26]. Therefore, SS was negatively correlated with SVA in our study.
The correlations of LMA at different intervertebral levels and severity of ADS were analyzed in our study. Results showed that U-LMA and L-LMA on the concave side were positively correlated with CA; U-LMA and L-LMA on the convex side were negatively correlated with CA. Yagi et al. [7] suggested that ADS patients did not have an age-related, progressive global muscle weakness, but rather a localized myopathy, which was commonly seen in patients with dropped head syndrome. Significant asymmetric LMA may be the primary cause of ADS and could be used to predict the progression of ADS. Convex U-LMA and concave L-LMA were negatively correlated with the SVA. This indicated that LMA on the diagonal through the apical vertebra would balance the stress and secure the stress conduction via parallelogram effect, which was very important to maintain sagittal imbalance. This parallelogram effect is first proposed in this study. Considering sagittal imbalance has a more disastrous influence on the clinical outcome than coronal imbalance in ADS patients, the significance of CVA may be limited [24]. R 2 value of multiple linear regression analysis of LMA and CVA was very low in our study, which indicated that LMA could not accurately predict CVA. This might be explained by the effects of various factors on CVA [27].  Schwab et al. [28] reported that no significant correlation was found between adult scoliosis and nutritional status or VAS scores in elderly patients. Similarly, our results showed that none of the parameters were correlated with VAS score. Therefore, VAS score would not be an effective method to measure the severity of ADS. In our study, PI was positively correlated with ODI; however, PT and SS were negatively correlated with ODI. Mac-Thiong et al. [29] found that sagittal spinal balance was strongly correlated with ODI in ADS; however, coronal spinal balance did not influence the ODI in their study. Therefore, influencing factors of SVA and ODI may be overlapping. Results in our study showed, at the lower intervertebral level, LMA on the concave side was positively correlated with ODI; in contrast, LMA on the convex side of ODI was negatively correlated with ODI. These indicated LMA at the lower intervertebral level would have a predictive effect on ODI in ADS patients. Compared with ODI, JOA scores are more focused on evaluating neurologic functions [30]. It was reported that worse sagittal spinopelvic alignment was the main cause of functional loss in ADS patients [20]. Therefore, JOA score seemed to be more correlated with spinopelvic parameters than LMA in our study.
There are several limitations associated with our study. First, this is a retrospective cross-sectional study that may result in unavoidable selection bias. Secondly, this is a single-center study and sample size is thus limited. Third, a comparative analysis of different phases of ADS was not conducted as the course ADS is difficult to follow. Studies with a larger sample size that include patients at different stages of disease progression are warranted to confirm the results of the present study.

Conclusion
In ADS patients, LMA on the concave side is more severe than the convex side; on the convex side, there are differences in the causes of U-LMA and L-LMA. LL and PT are both protective factors of CA progression; high LL is a risk of sagittal imbalance. High SS is a protective factor of sagittal imbalance. Asymmetric LMA may be positively correlated with CA. LMA on the diagonal through the apical vertebra may be very important to maintain sagittal imbalance via parallelogram effect. In addition, LMA at lower intervertebral levels of the apical vertebra may have a predictive effect on ODI. JOA score seems to be more correlated with spinopelvic parameters than LMA.